ir3: Support variants with different constlen's
[mesa.git] / src / freedreno / ir3 / ir3_compiler.h
1 /*
2 * Copyright (C) 2013 Rob Clark <robclark@freedesktop.org>
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
10 *
11 * The above copyright notice and this permission notice (including the next
12 * paragraph) shall be included in all copies or substantial portions of the
13 * Software.
14 *
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21 * SOFTWARE.
22 *
23 * Authors:
24 * Rob Clark <robclark@freedesktop.org>
25 */
26
27 #ifndef IR3_COMPILER_H_
28 #define IR3_COMPILER_H_
29
30 #include "ir3.h"
31
32 struct ir3_ra_reg_set;
33 struct ir3_shader;
34
35 struct ir3_compiler {
36 struct fd_device *dev;
37 uint32_t gpu_id;
38 struct ir3_ra_reg_set *set;
39 struct ir3_ra_reg_set *mergedregs_set;
40 uint32_t shader_count;
41
42 /*
43 * Configuration options for things that are handled differently on
44 * different generations:
45 */
46
47 /* a4xx (and later) drops SP_FS_FLAT_SHAD_MODE_REG_* for flat-interpolate
48 * so we need to use ldlv.u32 to load the varying directly:
49 */
50 bool flat_bypass;
51
52 /* on a3xx, we need to add one to # of array levels:
53 */
54 bool levels_add_one;
55
56 /* on a3xx, we need to scale up integer coords for isaml based
57 * on LoD:
58 */
59 bool unminify_coords;
60
61 /* on a3xx do txf_ms w/ isaml and scaled coords: */
62 bool txf_ms_with_isaml;
63
64 /* on a4xx, for array textures we need to add 0.5 to the array
65 * index coordinate:
66 */
67 bool array_index_add_half;
68
69 /* on a6xx, rewrite samgp to sequence of samgq0-3 in vertex shaders:
70 */
71 bool samgq_workaround;
72
73 /* The maximum number of constants, in vec4's, across the entire graphics
74 * pipeline.
75 */
76 uint16_t max_const_pipeline;
77
78 /* The maximum number of constants, in vec4's, for VS+HS+DS+GS. */
79 uint16_t max_const_geom;
80
81 /* The maximum number of constants, in vec4's, for FS. */
82 uint16_t max_const_frag;
83
84 /* A "safe" max constlen that can be applied to each shader in the
85 * pipeline which we guarantee will never exceed any combined limits.
86 */
87 uint16_t max_const_safe;
88
89 /* The maximum number of constants, in vec4's, for compute shaders. */
90 uint16_t max_const_compute;
91
92 /* on a3xx, the unit of indirect const load is higher than later gens (in
93 * vec4 units):
94 */
95 uint32_t const_upload_unit;
96 };
97
98 struct ir3_compiler * ir3_compiler_create(struct fd_device *dev, uint32_t gpu_id);
99
100 int ir3_compile_shader_nir(struct ir3_compiler *compiler,
101 struct ir3_shader_variant *so);
102
103 /* gpu pointer size in units of 32bit registers/slots */
104 static inline
105 unsigned ir3_pointer_size(struct ir3_compiler *compiler)
106 {
107 return (compiler->gpu_id >= 500) ? 2 : 1;
108 }
109
110 enum ir3_shader_debug {
111 IR3_DBG_SHADER_VS = BITFIELD_BIT(0),
112 IR3_DBG_SHADER_TCS = BITFIELD_BIT(1),
113 IR3_DBG_SHADER_TES = BITFIELD_BIT(2),
114 IR3_DBG_SHADER_GS = BITFIELD_BIT(3),
115 IR3_DBG_SHADER_FS = BITFIELD_BIT(4),
116 IR3_DBG_SHADER_CS = BITFIELD_BIT(5),
117 IR3_DBG_DISASM = BITFIELD_BIT(6),
118 IR3_DBG_OPTMSGS = BITFIELD_BIT(7),
119 IR3_DBG_FORCES2EN = BITFIELD_BIT(8),
120 IR3_DBG_NOUBOOPT = BITFIELD_BIT(9),
121 IR3_DBG_NOFP16 = BITFIELD_BIT(10),
122
123 /* DEBUG-only options: */
124 IR3_DBG_SCHEDMSGS = BITFIELD_BIT(20),
125 IR3_DBG_RAMSGS = BITFIELD_BIT(21),
126 };
127
128 extern enum ir3_shader_debug ir3_shader_debug;
129
130 static inline bool
131 shader_debug_enabled(gl_shader_stage type)
132 {
133 if (ir3_shader_debug & IR3_DBG_DISASM)
134 return true;
135
136 switch (type) {
137 case MESA_SHADER_VERTEX: return !!(ir3_shader_debug & IR3_DBG_SHADER_VS);
138 case MESA_SHADER_TESS_CTRL: return !!(ir3_shader_debug & IR3_DBG_SHADER_TCS);
139 case MESA_SHADER_TESS_EVAL: return !!(ir3_shader_debug & IR3_DBG_SHADER_TES);
140 case MESA_SHADER_GEOMETRY: return !!(ir3_shader_debug & IR3_DBG_SHADER_GS);
141 case MESA_SHADER_FRAGMENT: return !!(ir3_shader_debug & IR3_DBG_SHADER_FS);
142 case MESA_SHADER_COMPUTE: return !!(ir3_shader_debug & IR3_DBG_SHADER_CS);
143 default:
144 debug_assert(0);
145 return false;
146 }
147 }
148
149 static inline void
150 ir3_debug_print(struct ir3 *ir, const char *when)
151 {
152 if (ir3_shader_debug & IR3_DBG_OPTMSGS) {
153 printf("%s:\n", when);
154 ir3_print(ir);
155 }
156 }
157
158 #endif /* IR3_COMPILER_H_ */